Journal
PLOS ONE
Volume 17, Issue 6, Pages -Publisher
PUBLIC LIBRARY SCIENCE
DOI: 10.1371/journal.pone.0267714
Keywords
-
Categories
Funding
- South African Medical Research Council (SAMRC) through its Division of Research Capacity Development
- South African National Treasury
Ask authors/readers for more resources
The current research on predicting Gleason grade groups using DNA sequences as the sole input source still faces challenges, but in the new problem of predicting tumor DNA sequences from matched-normal DNA sequences, machine learning models showed improved performance.
One of the most precise methods to detect prostate cancer is by evaluation of a stained biopsy by a pathologist under a microscope. Regions of the tissue are assessed and graded according to the observed histological pattern. However, this is not only laborious, but also relies on the experience of the pathologist and tends to suffer from the lack of reproducibility of biopsy outcomes across pathologists. As a result, computational approaches are being sought and machine learning has been gaining momentum in the prediction of the Gleason grade group. To date, machine learning literature has addressed this problem by using features from magnetic resonance imaging images, whole slide images, tissue microarrays, gene expression data, and clinical features. However, there is a gap with regards to predicting the Gleason grade group using DNA sequences as the only input source to the machine learning models. In this work, using whole genome sequence data from South African prostate cancer patients, an application of machine learning and biological experiments were combined to understand the challenges that are associated with the prediction of the Gleason grade group. A series of machine learning binary classifiers (XGBoost, LSTM, GRU, LR, RF) were created only relying on DNA sequences input features. All the models were not able to adequately discriminate between the DNA sequences of the studied Gleason grade groups (Gleason grade group 1 and 5). However, the models were further evaluated in the prediction of tumor DNA sequences from matched-normal DNA sequences, given DNA sequences as the only input source. In this new problem, the models performed acceptably better than before with the XGBoost model achieving the highest accuracy of 74 +/- 01, F1 score of 79 +/- 01, recall of 99 +/- 0.0, and precision of 66 +/- 0.1.
Authors
I am an author on this paper
Click your name to claim this paper and add it to your profile.
Reviews
Recommended
No Data Available